Amazon's AI Chatbot 'Q' Struggles With 'Hallucinations' Amid Criticism For Rushed Launch

Comments
Loading...

The newly launched AI chatbot from Amazon.com Inc AMZN, dubbed Q, has been under fire from users and company insiders alike for producing incorrect outputs, colloquially known as ‘hallucinations.’

What Happened: Insiders at Amazon have linked these performance issues to a hurried launch and the deployment of a less proficient version of Anthropic’s Claude AI model. In response, Amazon is expanding its team of human reviewers to manually fine-tune the chatbot’s outputs, reported Business Insider on Wednesday.

Q, despite being Amazon’s most prominent generative AI product, was launched later than its rivals Microsoft Corp MSFT, Alphabet Inc’s Google GOOG GOOGL, and OpenAI. Amazon now faces the challenge of improving Q’s performance to stay competitive in the generative AI industry.

An Amazon spokesperson defended the launch strategy and emphasized that Q uses multiple AI models via Amazon Bedrock. They also noted the considerable positive feedback received from customers during the preview period.

See Also: Home Depot Goes From Home Repair, Appliances To Making Consumer Homes Smart

The deployment of the Claude AI model aligns with Amazon’s investment strategy, with up to $4 billion poured into the AI startup Anthropic. However, some insiders suggest upgrading to a superior version of Claude could boost Q’s capabilities.

Despite concerns raised by Q’s hallucinations, Amazon is beefing up its human evaluation process, also known as “human in the loop.” A dedicated task force is working on addressing these problems.

The efforts to enhance Q, however, are hampered by internal competition for AWS’s computing capabilities, especially with increased demand for GPUs from NVIDIA Corp NVDA and other suppliers.

Q’s launch is a part of Amazon’s 3-stack AI strategy, which includes user apps built on AI language models, the large language models themselves, and the computing power and chips.

Read Next: Apple’s Vision Pro Vs. Meta’s Quest 3: Virtual Reality Showdown

Image Via Shutterstock


Engineered by Benzinga Neuro, Edited by Kaustubh Bagalkote


The GPT-4-based Benzinga Neuro content generation system exploits the extensive Benzinga Ecosystem, including native data, APIs, and more to create comprehensive and timely stories for you. Learn more.


Market News and Data brought to you by Benzinga APIs

Posted In:
Benzinga simplifies the market for smarter investing

Trade confidently with insights and alerts from analyst ratings, free reports and breaking news that affects the stocks you care about.

Join Now: Free!